基于^(18)F-FDG PET/CT影像组学构建术前预测卵巢上皮癌淋巴结转移风险模型研究

A PET/CT-based radiomics nomogram for preoperative predicting lymph node metastases in epithelial ovarian cancer

目的:旨在结合临床特征构建基于PET/CT的影像组学模型,预测卵巢上皮癌淋巴结转移风险。方法:回顾性分析2017年1月—2020年11月接受肿瘤细胞减灭术的卵巢上皮癌患者275例,按照7∶3的比例随机划分为训练组(189例)和验证组(86例)。由两名经验丰富的放射科医生进行手工感兴趣区的勾画并提取组学特征。单变量分析筛选组学特征,Lasso回归分析分别构建基于PET及CT图像组学特征的组学标签。通过多因素Logistic回归分析构建模型,共构建三个模型:临床模型,临床+PET标签模型,临床+CT标签模型。通过受试者工作曲线、校准曲线及净重分类指数对上述模型进行评估,最佳模型构建列线图。结果:Lasso回归分析建立了PET及CT的组学标签,训练组中临床特征+CT组学标签模型的预测能力最好,AUC为0.726(95%CI 0.654~0.797),敏感度0.705,特异度0.667。而在验证组中临床特征+PET组学标签模型预测能力最好,AUC为0.713(95%CI 0.602~0.825),敏感度0.750,特异度为0.652,远高于其他模型。三个模型校准图显示临床特征+PET组学标签的模型在训练组和验证组中均有良好的校准度。临床特征+PET组学标签的模型泛化能力可能更强,我们为这个模型构建了列线图。结论:与单独使用临床特征相比,整合PET/CT的影像组学标签至模型中,可以提高模型诊断准确性并提供互补信息。在研究中构建的影像组学模型列线图是预测卵巢上皮癌患者淋巴结转移风险及高级别浆液性癌无进展生存的有效工具。

Objective:To construct radiomics model based on PET/CT images integrated with clinical features for prediction of lymph node metastasis in epithelial ovarian cancer.Methods:A total of 275 patients with epithelial ovarian cancer who underwent cytoreductive surgery from January 2017 to November 2020 were retrospectively analyzed.They were randomly divided into training group(n=189)and validation group(n=86).The ROIs were drawn by two experienced radiologists.The radiomics features were extracted.Univariate analysis was used to screen the omics features,and Lasso regression analysis was used to construct the omics labels based on the omics features of PET and CT images.Three models were constructed by multivariate Logistic regression analysis(clinical,clinical+PET,and clinical+CT).Models were evaluated by the receiver operating curves,calibration curves,and net reclassification index.The best model was constructed into a nomogram.Results:Lasso regression analysis was used to construct the omics labels of PET and CT.In the training group,the prediction ability of clinical characteristics+CT omics label model was the best.The AUC,sensitivity,and specificity of this model was 0.726(95%CI0.654~0.797),0.705,and 0.667.In the validation group,the prediction ability of clinical features+PET omics label model was the best,and the AUC,sensitivity,and specificity was 0.713(95%CI 0.602~0.825),0.750,and 0.652,which is much higher than other models.The three model calibration charts showed that the model of clinical features+PET omics label had good calibration in the training group and validation group.The model generalization ability of clinical features+PET omics label was stronger,so we constructed a nomogram for this model.Conclusion:Compared with using clinical features alone,integrating PET/CT omics labels into the model can improve the diagnostic accuracy of the model and provide complementary information.The radiomics model nomogram constructed in this study is an effective tool to predict the risk of lymph node metastasis and progression free survival of high-grade serous cancer in patients with epithelial ovarian cancer.